A data-driven approach to predicting critical TSP episodes: insights from a mining-influenced area in Southeast Brazil
Year: 2025
Type: article
Abstract: Air pollution caused by Total Suspended Particles (TSP) poses significant environmental and health risks, particularly in regions with intensive soil-disturbing activities, such as mining areas. TSP concentrations are influenced by complex and nonlinear interactions between meteorological conditions and surface-related factors, presenting substantial challenges for controlling TSP emissions and their associated impacts. Few studies have developed data-driven approaches to identify and anticipate high-risk periods based on the dynamics of these drivers. This study aims to characterise the temporal dynamics and key drivers of TSP, and to delineate thresholds of these drivers associated with critical TSP episodes in a mining-intensive region of southeastern Brazil. Using observational data from 2023 and applying Kernel Ridge Regression (KRR), we explored hourly, daily, and monthly TSP patterns, assessed meteorological and surface-related drivers, and delineated thresholds of key drivers associated with extreme pollution episodes. Results show that TSP levels peak during dry months (July–September), on weekdays, and during morning and evening rush hours. Strong associations were observed between TSP and PM₁₀, with dry and low-humidity conditions favouring the increase of these pollutants. The KRR model explained 71% of the TSP variability, highlighting soil moisture, wind gusts, and consecutive dry periods as key predictors. Thresholds such as soil moisture <0.15 m3/m3, wind gusts >16 m/s, and consecutive dry periods >470 hours were critical for extreme events. This study provides a framework for anticipating critical TSP episodes based on key meteorological and surface-related drivers, enabling informed mitigation strategies in mining-impacted regions.
Source: Figshare
Authors: Ribeiro, Camilo Bastos, Hoinaski, Leonardo, Will, Robson, Hess, José Henrique, Ribeiro, Amanda, Pedruzzi, Rizzieri, de Freitas, Edmilson Dias, de Almeida Albuquerque, Taciana Toledo
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7084044002
FWCI:
Cites:
Cited by:
Related to: